Multicommodity Rent or Buy: Approximation Via Cost Sharing Martin Pál Joint work with Anupam Gupta...

Multicommodity Rent or Buy:Approximation Via Cost Sharing

Martin Pál

Joint work withAnupam Gupta Amit Kumar

Tim Roughgarden

•Cost sharing algorithm for Steiner forest: what do we want

•Cost sharing for Steiner forest +Simple randomized analysis = Approximation for Rent or Buy

•How does the cost sharing work

Talk Outline

Steiner Forest and Rent or Buy

Input:

•Weighted graph G

•Set of demand pairs D

Solution:

A set of paths, one for each (si,ti) pair

Steiner Forest and Rent or Buy

Input:

•Weighted graph G

•Set of demand pairs D

Steiner Forest: pay ce for each edge e we use, regardless of how many paths use it

Steiner Forest and Rent or Buy

Input:

•Weighted graph G

•Set of demand pairs D

Rent or Buy: pay ce per pair for renting, or buy for Mce

Related Work

•2-approximation for Steiner Forest[Agarwal, Klein, Ravi 91] , [Goemans, Williamson 95]

•Constant approximation for Rent or Buy[Kumar, Gupta, Roughgarden 02]

•Spanning Tree + Randomization = Single Sink Rent or Buy[Gupta, Kumar, Roughgarden 03]

Cost Sharing for Steiner Forest

Cost sharing algorithm: approximation algorithm that given D, computes solution FD and set of cost shares (D,j) for jD

Cost Sharing for Steiner Forest

Cost sharing algorithm: approximation algorithm that given D, computes solution FD and set of cost shares (D,j) for jD

(P1) Constant approximation: cost(FD) St*(D)

(P2) Cost shares do not overpay: j (D,j) St*(D)

(P3) Cost shares pay enough: let D’ = D {si, ti}

dist(si, ti) in G/ FD (D’,i)

What does (P3) say

What does (P3) say

Solution for Solution for

a

c

( , ) (a+c)

b

The Rent or Buy Algorithm

Algorithm SimpleMRoB:

1. Mark each demand pair independently at random with prob. 1/M. Let S be the marked set.

2. Use a cost sharing algorithm to build a Steiner forest on the marked set S.

3. Rent shortest paths between all unmarked pairs.

Analysis

Claim: Expected cost of buying an optimal Steiner forest on S is at most OPT.

Corollary: Expected cost of step 2 OPT

Corollary: Expected sum of cost shares OPT

Analysis (2)

Claim: Expected cost of step 3 is at most OPT.

Let S’ = S – {i}

E[cost share of i | S’] = 1/M M (S’+i, i)

E[rental cost of i | S’] = (M-1)/M (dist(sj, tj) in G/FS’ )

From (P3):

E[rental cost of i | S’] E[cost share of i | S’]

E[rental cost of i] E[cost share of i]

Analysis (3)

There is a cost sharing algorithm with = 6, = 6.

Theorem: SimpleMRoB is a 12-approximation algorithm.

Existing Steiner forest algorithms

•Each demand starts in a separate cluster

•Active clusters grow at unit rate

•When two clusters touch, they merge into one

•Demands may get deactivated

•A cluster is deactivated if it has no active demands

•Inactive clusters do not grow

•Keep adding enough edges so that all active demands in a component are connected

Existing Steiner forest algorithms

Existing Steiner forest algorithms

Existing Steiner forest algorithms

Existing Steiner forest algorithms

Existing Steiner forest algorithms

Existing Steiner forest algorithms

How to define cost shares

Need to pay for the growth of the clusters

Active demands share equally the cost of growth of a cluster.

a(u,) = 1 / (# of active demands in cluster with u)

= 0 if u not active at time

cost share of u = a(u,) d

Does it work?

(P1) Constant approximation: cost(FD) St*(D)

(P2) Cost shares do not overpay: j (D,j) St*(D)

(P3) Cost shares pay enough: let D’ = D {si, ti}

dist(si, ti) in G/ FD (D’,i)

(P1) and (P2) easy to verify. How about (P3)?

Does it work?

2 22+

Does it work?

2 22+

Does it work?

2 22+

cost share of = 2/n +

Does it work?

2 22+

cost share of = 2/n + << 2 +

solution without

Solution

Inflate the balls !

1. Run the standard [AKR, GW] algorithm

2. Note the time Tj when each demand j deactivated

3. Run the algorithm again, except that now every demand j deactivated at time Tj for some > 1

Adopting proof from [GW]: buying cost at most 2OPT.

Works on our example

= 2

Works on our example

Works on our example

Works on our example

How to prove (P3)

Need to compare runs on D and D’ = D {si, ti}

Idea: 1. pick a {si, ti} path P in G/ FD

2. Show that (D’,i) accounts for a constant fraction of length(P)

How to prove (P3)

Need to compare runs on D and D’ = D {si, ti}

Idea: 1. pick a {si, ti} path P in G/ FD

2. Show that (D’,i) accounts for a constant fraction of length(P)

Instead of (D’,i), use alone(i), the total time si or ti was alone in its cluster

How to prove (P3)

Need to compare runs on D and D’ = D {si, ti}

Idea: 1. pick a {si, ti} path P in G/ FD

2. Show that (D’,i) accounts for a constant fraction of length(P)

How to prove (P3)

Need to compare runs on D and D’ = D {si, ti}

Idea: 1. pick a {si, ti} path P in G/ FD

2. Show that (D’,i) accounts for a constant fraction of length(P)

How to prove (P3)

Need to compare runs on D and D’ = D {si, ti}

Idea: 1. pick a {si, ti} path P in G/ FD

2. Show that (D’,i) accounts for a constant fraction of length(P)

(D’, ), = 2.5

alone( ) = 2.0

Mapping of layers

Lonely layer: generated by si or ti while the only active demand in its cluster

Each non-lonely layer in the D’ run maps to layers in the scaled D run.

Idea:

length(P) = #of lonely and non-lonely layers crossed in D’ run

length(P) #of layers it crosses in scaled D run

Hence: for each non-lonely layer, -1 lonely layers crossed.

Mapping not one to one

Two non-lonely layers in D’ run can map to the same layer in scaled D run.

D’ run

scaled D run

Not all layers of D run cross P

A layer in D’ run that crosses P can map to a layer in scaled D run that does not cross P

D’ run scaled D runsi si

The “waste” can be bounded.

Open problems

• Does SimpleMRoB work with unscaled GW? With an arbitrary Steiner forest subroutine?

• Other cost sharing functions? (crossmonotonic..)